Speaker: Andrew Parnell, Associate Professor, School of Mathematics and Statistics, University College Dublin
Title: Bayesian Additive Regression Trees: Bridging Machine Learning and Statistics
Abstract: Regression trees have proven to be one of the most useful methods in statistics and machine learning as they naturally capture complex non-linear relationships between variables and have consistently shown excellent out-of-sample performance. A number of extensions, the most popular of which is Random Forests, have extended the idea and have become part of the standard machine learning toolbox. In this talk I will provide an introduction and history of tree-based methods, and outline a more recent version, Bayesian Additive Regression Trees (BART). I will discuss some of the recent extensions we have built on to BART, and show how they perform in real world examples.